Plan for tomorrow's FESCo meeting (2010-11-17)

[Kevin Kofler]

Adam Williamson wrote:
> But the fact remains that *right now*, this is what Fedora is. I think
> that it makes sense to commit to being whatever we are fully. Right now,
> we're a stable release distribution; we should work to make those
> releases properly stable, to actually be what we represent ourselves as
> being. If we follow your strategy of pushing as many updates as possible
> as aggressively as possible to all stable releases, we're not being a
> stable release distribution at all, we're just shipping three branches
> at a time which are all essentially rolling releases and have absolutely
> no guarantee of stability. It's the worst of all worlds: we get all the
> messy overheads of supporting three releases at a time, but none of the
> benefits you get from properly following a stable release model.

Well, it's here that we disagree. For me, it's actually the best of both 
worlds:
* as for the strict stable release model and unlike the rolling release 
model, you get to avoid those changes most likely to break your setup or 
require manual intervention by staying on the stable release, because 
maintainers should avoid pushing such disruptive changes to the stable 
release,
* as for the rolling release model and unlike the strict stable release 
model, you benefit from incremental, backwards-compatible improvements as 
soon as upstream releases them,
so in short, you get all the cool new stuff without the breakage.

FWIW, in theory, this isn't even so different from the stable updates vision 
voted by the Board: they also define a set of disruptive update types which 
should not be pushed to stable releases; the big difference is that their 
idea of what's "disruptive" is VERY DIFFERENT from mine, and IMHO is 
throwing out the baby with the bathwater (it prevents even those updates 
that ARE wanted). In addition, I don't like the wording which essentially 
says "don't push an upgrade unless there's a reason to", I believe we should 
say "always push an upgrade unless there's a reason NOT to".

> It seems like what you want is actually not to have three releases at a
> time at all but to have one and update it constantly. And I actually
> rather suspect that would be a model that would work well for Fedora,
> and I'd like to look into adopting it. But it's *not* the model we have
> right now, and it doesn't strike me as a bad idea to try and abuse a
> stable release model by pushing admittedly unstable changes into the
> stable releases. A stable release into which we cheerily shove new
> versions of everything just for the hell of it really isn't a stable
> release, and it has no clear identity separation from the next or
> previous stable releases, and hence makes the whole idea of having
> releases rather pointless and just dead weight overhead.

No, I don't actually want that. You're attacking a strawman. Stop putting 
words in my mouth!

I DON'T want to get an upgrade such as the one from KDE 3 to 4, the one from 
Amarok 1 to 2, the one from KDevelop 3 to 4, the one from GNOME 2 to 3 etc. 
as a regular update! Those are what new releases are for! (And there's your 
"clear identity separation".) Minor feature release upgrades such as from 
Amarok 2.2 to 2.3 (which DON'T have feature regressions, configuration 
incompatibilites etc.), on the other hand, are perfectly fine to push as an 
update, since they're clear improvements (and actually also fix bugs in 
addition to adding features).

        Kevin Kofler